d/numam-dpdk
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
17cbf09fe1
numam-dpdk/lib/librte_sched
History
Stephen Hemminger 03d00293ca sched: eliminate floating point in calculating byte clock 
The old code was doing a floating point divide for each rte_dequeue()
which is very expensive. Change to using fixed point scaled inverse
multiply. To maintain equivalent precision, scaled math is used.
The application ABI is the same.

This improved performance from 5Gbit/sec to 10 Gbit/sec when configured
for 10 Gbit/sec rate.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
2016-03-13 23:31:59 +01:00
..
Makefile sched: introduce reciprocal divide 2016-03-13 23:31:59 +01:00
rte_approx.c remove extra parentheses in return statement 2016-02-10 15:47:50 +01:00
rte_approx.h remove trailing whitespaces 2014-06-11 00:29:34 +02:00
rte_bitmap.h bitmap: optimize for 128-bytes cache line 2016-02-11 12:45:35 +01:00
rte_reciprocal.c sched: introduce reciprocal divide 2016-03-13 23:31:59 +01:00
rte_reciprocal.h sched: introduce reciprocal divide 2016-03-13 23:31:59 +01:00
rte_red.c remove trailing whitespaces 2014-06-11 00:29:34 +02:00
rte_red.h remove extra parentheses in return statement 2016-02-10 15:47:50 +01:00
rte_sched_common.h remove trailing whitespaces 2014-06-11 00:29:34 +02:00
rte_sched_version.map sched: hide structure of port hierarchy 2015-07-17 00:35:49 +02:00
rte_sched.c sched: eliminate floating point in calculating byte clock 2016-03-13 23:31:59 +01:00
rte_sched.h sched: keep track of RED drops 2016-03-13 23:28:00 +01:00